OBJECTIVE: To establish an analysis pipeline for the volumetric evaluation of the osteotomy site after bilateral sagittal split osteotomy (BSSO).
METHODS: Cone-beam computed tomography (CBCT) was performed before, directly after BSSO, and 6-12 months after surgery. Image segmentations of each osteotomy gap data set were performed manually by four physicians and were compared to a semi-automatic segmentation approach.
RESULTS: Five patients with a total of ten osteotomy gaps were included. The mean interclass correlation coefficient (ICC) of individual patients was 0.782 and the standard deviation 0.080 when using the manual segmentation approach. However, the mean ICC of the evaluation of anatomical sites and time points separately was 0.214, suggesting a large range of deviation within the manual segmentation of each rater. The standard deviation was 0.355, further highlighting the extent of the variation. In contrast, the semi-automatic approach had a mean ICC of 0.491 and a standard deviation of 0.365, which suggests a relatively higher agreement among the operators compared to the manual segmentation approach. Furthermore, the volume of the osteotomy gap in the semi-automatic approach showed the same tendency in every site as the manual segmentation approach, but with less deviation.
CONCLUSIONS: The semi-automatic approach developed in the present study proved to be valid as a standardised method with high repeatability. Such image analysis methods could help to quantify the progression of bone healing after BSSO and beyond, eventually facilitating the earlier identification of patients with retarded healing.

UNASSIGNED: Segmentation of oropharyngeal squamous cell carcinoma (OPSCC) is needed for radiotherapy planning. We aimed to segment the primary tumor for OPSCC on MRI using convolutional neural networks (CNNs). We investigated the effect of multiple MRI sequences as input and we proposed a semi-automatic approach for tumor segmentation that is expected to save time in the clinic.
UNASSIGNED: We included 171 OPSCC patients retrospectively from 2010 until 2015. For all patients the following MRI sequences were available: T1-weighted, T2-weighted and 3D T1-weighted after gadolinium injection. We trained a 3D UNet using the entire images and images with reduced context, considering only information within clipboxes around the tumor. We compared the performance using different combinations of MRI sequences as input. Finally, a semi-automatic approach by two human observers defining clipboxes around the tumor was tested. Segmentation performance was measured with Sørensen-Dice coefficient (Dice), 95th Hausdorff distance (HD) and Mean Surface Distance (MSD).
UNASSIGNED: The 3D UNet trained with full context and all sequences as input yielded a median Dice of 0.55, HD of 8.7 mm and MSD of 2.7 mm. Combining all MRI sequences was better than using single sequences. The semi-automatic approach with all sequences as input yielded significantly better performance (p < 0.001): a median Dice of 0.74, HD of 4.6 mm and MSD of 1.2 mm.
UNASSIGNED: Reducing the amount of context around the tumor and combining multiple MRI sequences improved the segmentation performance. A semi-automatic approach was accurate and clinically feasible.